A/B-Test with 2 different designs, server-sided[ Edited ]
October 2016 - last edited October 2016
we are discussing 2 different web-designs for a webpage. The problem is that the turnover is highly weather-depended, so A/B-tests like "Show version A for 7 days then show version B for 7 days" seem not to work to get a good comparisson.
So we are thinking about serving the 2 versions "at the same time", meaning one user gets version A and another user some minutes later gets version B. The problem is that Analytics seem to make experiments available only for different urls (using redirects) or on client-side changes (https://developers.google.com/analytics/solutions/experiments-client-side).
But what we need would be a "server-side experiment", because the decission, which template to serve, has to be made on the server. So the idea was to randomly decided which design should be serverd to the user, and to set a cookie so a user which is navigating through the side or coming back after some time is not suddenly getting the other design.
The question is: Is it possible to use analytics-experiments for this as well? Would there be any advantage compared to using lets say just a user-defined dimension for the design and checking the turnover for both designs?